Breaking the Chain: discover the reasoning limits of Large Language Models (LLMs) using a simple and inexpensive Word Swap Challenge. Most models, even advanced ones, struggle with basic sequential reasoning tasks after just two operations!

The GoodAI Grants program has awarded @tudelft University w/a hardware endowment for their AI research! We're especially excited about Assistant Prof. Dr. @WendelinBoehmer's research on Generalization in Distributed Badger Architectures! Read on for more: goodai.com/goodai-grants-pro…

(1/2) Thank you everyone who joined us for the Badger Seminar over the last 5 days both online & in person! It has been hugely productive for us to connect over the topic of Beyond Life-long Learning via Modular Meta-Learning & discuss in small groups...

Is there a way to move beyond reward functions? Can we make agents that "try" to do things, in any way other than telling them to maximise something? Some of the questions being discussed this morning at the Badger Seminar.

What are we missing to get an AI that has common sense? Do we just need to maximize the right metric, or do we need a completely new mechanism/design? Some of the questions being discussed this morning at the Badger Seminar online and in Prague.

(1/2) Interesting questions from the Badger Seminars. Humans are capable of on-the-fly adaptation - we learn many simple tasks quickly, by combining previous skills, much more than specialize in a specific area...

Another fascinating discussion from the Badger Seminars with @36zimmer, @okw, @NathanielVirgo. Given agents with sufficiently powerful world models, what do we need to add to get memes / memetic evolution among them?

Ice breaker this morning at the Badger Seminar! Can anyone figure out what they are trying to draw?

The first day of the Badger Seminar has kicked off! With over 30 researchers online & in Prague! The topic for the first day is: Learning to learn; Lifelong learning (continual learning and gradual learning).

Day 4 of the Badger Seminar is underway! Today we are discussing “Manual Badger” vs “Automatic Badger” approach, learn more about the approaches in our Research Roadmap: goodai.com/goodai-research-r…

Another great day of discussions with the GoodAI Research team, @okw, @36zimmer, @NathanielVirgo, @blake_camp_1, and many more! Today we were Exploring the Manual and Automatic Badger approaches. Looking forward to the final day tomorrow!

We have released the GoodAI Research Roadmap for 2021/2022. It focuses on 4 key areas: learning to learn, lifelong (gradual) learning, open-endedness, & generalization/extrapolation of meta-learned algorithms. Find out more. goodai.com/goodai-research-r…

"Building a research culture where we share questions & difficulties, not just answers." This is the goal of the GoodAI Grants. Find out more about that vision in this blog from @ngutten, and let us know what you think. goodai.com/building-a-resear…

Using open-ended algorithms to generate video game content in @SpaceEngineersG. We are excited to announce the latest GoodAI Grant collaboration with Dr. Kai Arulkumaran (@kaixhin) of Araya. Find out more. goodai.com/using-open-ended-…

GoodAI recently contributed to the @EU_Commission report on Humans & Society in the Age of AI. Find out about our contribution and read the full report from @VladimirSucha & @JPGammel . goodai.com/goodai-contribute…

"Solving generalization & making artificial intelligence curious." Happy to announce the latest GoodAI Grant awarded to Deepak Pathak (@deepakpathak) from CMU (@SCSatCMU). Addressing some shortcomings of modern AI. Find out more. Photo credit: Nitesh Mor. goodai.com/solving-generaliz…

Bayesian Online Meta-Learning (BOML) for continual & gradual learning. We are happy to announce our latest GoodAI Grant which has been awarded to Pauching Yap, PhD Student at the UCL Centre for Artificial Intelligence. @ai_ucl. Find out more. goodai.com/bayesian-online-m…

GoodAI Grants update! We announced our first 4 GoodAI Grants this week featuring, meta-learning and combinatorial generalization, multi-agent AI systems, open-endedness, and novel behaviors. Find out more in our newsletter. mailchi.mp/c68d19ea2689/good…